2.1 Borrowers’ states

Geographical information is critical for the growth of a company. Firstly, analyze what state owns the most of loans on Proser Marketplace. However, since the population density is vary from state to state, this result may be less informative than people normally think.

## 
##          AK    AL    AR    AZ    CA    CO    CT    DC    DE    FL    GA 
##  5515   200  1679   855  1901 14717  2210  1627   382   300  6720  5008 
##    HI    IA    ID    IL    IN    KS    KY    LA    MA    MD    ME    MI 
##   409   186   599  5921  2078  1062   983   954  2242  2821   101  3593 
##    MN    MO    MS    MT    NC    ND    NE    NH    NJ    NM    NV    NY 
##  2318  2615   787   330  3084    52   674   551  3097   472  1090  6729 
##    OH    OK    OR    PA    RI    SC    SD    TN    TX    UT    VA    VT 
##  4197   971  1817  2972   435  1122   189  1737  6842   877  3278   207 
##    WA    WI    WV    WY 
##  3048  1842   391   150

From the bar plot above, we can see that California has most total loans. Meanwhile, the total number of loans distributed similarly as population. One interesting find is there are 5515 loans has unkonwn state, are they from out of United States or it is data entry error? It is hard to allocate these loans to a specific states without additional data from Prosper Marketplace.

2.2 Occupation

Will loaner’s occupation affects there chance of getting investors? Is there any occupation discriminate on Prosper MarketPlace? To better answer these question, total amount of loans for different occupations are analyzed first.

Obviously, the dropdown menu on Prosper Marketplaces’s website is not exhaustive. The occupation “Other” siginificantly skewed the distribution.

2.3 Employment Status Duration

##  [1] 136 127   0  39  11  33  42 422 284 166   3  94  10  18   6

As shown in the graph above, most of the loaners provided employement data on this website do have a fairly short employment status history.

2.4 Credit Score

The credit score records is dvided into two features: 1. CreditScoreRangeLower 2. CreditScoreRangeUpper Firstly, the relationship between this two features is determined.

## 
##     19 
## 113346

## 
##    19   379   439   459   479   499   519   539   559   579   599   619 
##   133     1     5    36   141   346   554  1593  1474  1357  1125  3602 
##   639   659   679   699   719   739   759   779   799   819   839   859 
##  4172 12199 16366 16492 15471 12923  9267  6606  4624  2644  1409   567 
##   879   899 
##   212    27

As show above, the the creditscorerangeupper is mostly 19 more than the creditscorerangelower. According to the properties of expected values. Only one of these variables needs to analyzed.

As shown above, the distribution of credit score shows approximate normal distribution if we neglect outliers at 19.

2.5 Income Range

## 
##             $0      $1-24,999      $100,000+ $25,000-49,999 $50,000-74,999 
##            621           7274          17337          32192          31050 
## $75,000-99,999  Not displayed   Not employed 
##          16916           7741            806

As show in the graph above, the income range of lenders is a good representation of financial state of lenders. Quite a number of incomes is not displayed. But mostly the income range of lenders is concentrated below $50,000.

2.6 Total Credit Line

2.7 Loan Origination Date

The purpose of analyzing loan origination date is verify of the data

2.8 Loan amount

## 
##  4000 15000 10000  5000  2000  3000 25000 20000  1000  2500  7500  7000 
## 14333 12407 11106  6990  6067  5749  3630  3291  3206  2992  2975  2949 
##  6000  3500  8000 12000  9000 13000  1500  4500 
##  2869  2567  2442  1921  1695  1509  1507  1406

Most of the loan amount is on numbers that looks more “integer” like 4000, 15000 insteand 15232 or 4355. Because the Prosper MarketPlace doesn’t gurantee this amount will be fully funded, its make sense that most lenders put a approximate number to take the most from investers.

2.10 Number of investors

2.11 Loan status (Main Feature)

There are multiple status of loans per data dictionary:

Cancelled Chargedoff Completed Current Defaulted FinalPaymentInProgress Past due(Followed by buckets)

##  [1] "Cancelled"              "Chargedoff"            
##  [3] "Completed"              "Current"               
##  [5] "Defaulted"              "FinalPaymentInProgress"
##  [7] "Past Due (>120 days)"   "Past Due (1-15 days)"  
##  [9] "Past Due (16-30 days)"  "Past Due (31-60 days)" 
## [11] "Past Due (61-90 days)"  "Past Due (91-120 days)"

As seen in this graph, the mojority of loans is in their “Current” Status. Becuase the loan status is also the main feature of our analysis, we need to go one step further on.

The loans status feature has 12 levels:

##  [1] "Cancelled"              "Chargedoff"            
##  [3] "Completed"              "Current"               
##  [5] "Defaulted"              "FinalPaymentInProgress"
##  [7] "Past Due (>120 days)"   "Past Due (1-15 days)"  
##  [9] "Past Due (16-30 days)"  "Past Due (31-60 days)" 
## [11] "Past Due (61-90 days)"  "Past Due (91-120 days)"

To make the analsis easier in later stage, these levels will be divided into 2 groups: Good loans and Bad loans, in the other word, the deliquency state of loans. After read the wikipedia and investopedia page of these words, they are divided as follow:

• Good loans():
  • Cancelled
  • Completed
  • Current
  • FinalPaymentInProgress
  • Past Due (1-15 days)
  • Past Due (16-30 days)
• bad (deliquented) loans:
  • Chargedoff
  • Defaulted
  • Past Due (>120 days)
  • Past Due (31-60 days)
  • Past Due (61-90 days)
  • Past Due (91-120 days)

A new column called “DeliquencyState” is created based on the creteria above. Under this column, 1 represent loans with deliquencies and 0 represents loans without deliquencies.

## 
##   delinquent GoodStanding 
##        18006        95931

2.12 Univariate Analysis

3.1 Loan status (Deliquency State )

3.1.1 Deliquency State and Employment Status Duration

3.1.2 Deliquency State and Income Range

Although it is obvious that borrowers that has income range from 25,000 to 49,999 has most of the delinquent loans, they also has most of loans in general. A more reasonable feature in this kind case would be the percent of deliquent loans compare to the total loans in that income range.

3.1.3 Deliquency State and loan original amount

In contraray to the intuition, delinquent loans has lower loan amount compare to good standing loans. I was expecting delinquent has higher loan amount which makes them harder to payoff.

3.1.3 Deliquency State and employment duration

The boxplot, in this case, does not make too much sense because of significant amount of data points that out of quantiles from the mean. Because of that, I used the point plot to have visulize the employement duration.

3.2.1 Number of Investors and loan category

Because all listing categories are in numeric forms, it is necessary to map the english descriptvie name to them according to the data dictionary.

3.2.2 Number of Investors and credit score

Will higher credit score attract more investors? A scatter plot with trend line might be able answer this question.

Since it is barely possible for a person has credit score of 0, I only plotted loans with credit score above 0. Although the relationship between credit score and is approximately position, it is noticeable that there is a downward trend after credit score go over 780.

3.2.3 Number of Investors and credit score

Interestingly, weather the income verifiable does not have obvious influence on the number of investors.

3.3.1 Loan original amount and loan category

3.3.2 Pearson Correlation

From the correlation plot above, we found fairly strong correlations between Investors and loan amount, credit score and prosper score. These three variables will be explored deeper in next plot. However, it is also necessary to keep in mind that ProsperScore have high correlations of .4 with creditrange.

Talk about some of the relationships you observed in this part of the  investigation. How did the feature(s) of interest vary with other features in  the dataset?

1. For the deliquency of loans, the duration of employement and credit score seems have most obvious relationship with the deliquency state. Longer employement history and higher the credit score seems relevant to lesser deliquent loans.

2. As of the number of investors, the credit score and loan categories both have effects. The credit score has much more significant effects on the number compare to loan categories. The number of investors increased as the credit score goes up.

Did you observe any interesting relationships between the other features  (not the main feature(s) of interest)?

The debt consolidation has highest median for loan amount amoung all other categories. Medical and dental loans does not own a high loan amount. The reason might be the residents of United States heavily rely on insurance and law service for health coverage.

What was the strongest relationship you found?

The strongest I found is the relationship between credit score and the number of investor.

4.1 Credit score, prosperscore, loan original amount and number of investors

## 
## Call:
## lm(formula = Investors ~ LoanOriginalAmount + CreditScoreRangeUpper, 
##     data = loan_df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -266.38  -59.78   -7.54   40.92 1027.17 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           -1.405e+02  3.063e+00  -45.86   <2e-16 ***
## LoanOriginalAmount     5.368e-03  4.773e-05  112.46   <2e-16 ***
## CreditScoreRangeUpper  2.502e-01  4.489e-03   55.74   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 94.42 on 113343 degrees of freedom
##   (591 observations deleted due to missingness)
## Multiple R-squared:  0.1666, Adjusted R-squared:  0.1666 
## F-statistic: 1.133e+04 on 2 and 113343 DF,  p-value: < 2.2e-16

The amateur regression model here does not represent the actual relationship between scores, loan amount and the number invester. Instead, p value less than 0.05 shows credit scores and loan amount do have statistically significant impact on the number of investers. From the Q-Q plot, it can ben seen that the distribution of investors is not normal, which is also shwon in the univariate analysis. From the Leverage plot, there’s a outlier located on the left of the graph. T his group of points indicate that there are mulitple outliers while doing the regression. From the residual plot, it can be seen that as the number of investors goes large, its become harder and harder to predict the number of investors.

Talk about some of the relationships you observed in this part of the

investigation. Were there features that strengthened each other in terms of
looking at your feature(s) of interest?

Relationship between number of investors and credit scores, derived in bivariate analysis are used in conjuntion with the original loan amount in multivariate. The result is significant, these two variables can be used together to built a regression model to predict the number of investors.

Were there any interesting or surprising interactions between features?

Seems the prosper score, which is the socre assigned by prosper loan marketplace has strong correlation with the credit score assigned by credit union. This is either because the prosper loan market place assign their score based on credit score from credit union or people’s bevaior on the prosper market place reflect their general credit behavior.

OPTIONAL: Did you create any models with your dataset? Discuss the

strengths and limitations of your model.

I tried to build a model based as: Investors ~ LoanOriginalAmount + CreditScoreRangeUpper. However, some spricification for linear regression is break, since the purpose is not to build a successful but to identify the general trend between variables, I didn’t dived further onto this model.

Plot One

Even the deliquency status of loans in significant impacted by the credit score of borrowers, the monthly income also plays a role in the behavior of deliquency status. For this plot, it is obvious that lower credit score implies higher chance of deliquent loans. Borrowers with higher monthly income will have higher credit score and higher chance of pay out their loans.

Plot Two

From the Q-Q plot, it can ben seen that the distribution of investors is not normal, which is also shwon in the univariate analysis. From the Leverage plot, there’s a outlier located on the left of the graph. This group of points indicate that there are mulitple outliers while doing the regression. From the residual plot, it can be seen that as the number of investors goes large, its become harder and harder to predict the number of investors.

Plot Three

From this plot, we got a interesting finding that loans with higher loan original amount owns bettwe chance of good standing at the end. There couple theories for this behaviors. Firstly, this might because these loan has longer payment cycle so that they are still going through the pay back cycle. It coule also because borrowers borrowed a lot from loan companies tend to making better judgement of their financial situation.

7 Reference

https://yihui.name/knitr/options/

https://www.rstudio.com/wp-content/uploads/2015/03/ggplot2-cheatsheet.pdf

https://www.statmethods.net/stats/regression.html

https://stackoverflow.com/questions/41205489/different-fill-values-for-different-factors-in-ggplot2